[Decision] objectui execution half of the merged responsive retirement (objectstack#11027) — deleting the two dead implementations does NOT unblock the spec bump, and the retirement's "no other authorable carrier" is false here #7580

Activity

1. os-project-manager

   added

   needs-user-decision

   domain:uiobjectui ui stream: fix lands on the published library or apps — objectui execution seatobjectui ui stream: fix lands on the published library or apps — objectui execution seat

   on Sep 4, 2026
2. os-warren commented on Sep 4, 2026

   

   os-warren

   on Sep 4, 2026 – with Claude

   Collaborator

   More actions

   Maintainer ruling recorded — A: re-home the breakpoint vocabulary in objectui; BreakpointOrderMap retires; both halves in one PR. needs-user-decision → pm:queue in the same stroke.

   Director seat, summon 14 (session_01LsEjuNMPitCHwEfYftZ1um, GitHub os-warren). Provenance: maintainer, live PM chat, 2026-09-04, decision batch #26, verbatim 「2 · objectui#7580 同意」 on the presented recommendation A (fallback C; B refused on measurement). Readings this ruling rests on were re-verified on objectui origin/main e30ed15 and objectstack origin/main ebb0822 at 2026-09-04T03:49Z.

   The ruling

   1. BreakpointName is declared locally in @object-ui/types (packages/types/src/mobile.ts, where the spec import at line 19 and the re-export at line 32 stand today) — the renderer-side layout vocabulary lives with every other component schema. Its live readers stay as they are: packages/mobile (breakpoints.ts, useBreakpoint.ts, ResponsiveContainer.tsx, useResponsiveConfig.ts) and ResponsiveValue<T>.
   2. BreakpointColumnMap is declared locally in packages/layout (ResponsiveGrid.tsx, replacing the spec import at line 11); responsive-grid's authorable columns keeps its type and its resolver.
   3. BreakpointOrderMap retires — ResponsiveGrid.tsx:26 itself records it has no read point; the re-export at :32 goes with it.
   4. The two dead implementations (useResponsiveConfig in @object-ui/mobile, ResponsiveProtocol in @object-ui/core, both measured at zero callers on [finding] 响应式词汇的两个零消费表面:useResponsiveConfig 生产零调用者(page.components[].responsive 因此实际未生效)、MobileComponentConfig 零消费者 #4773) are deleted, and packages/mobile/src/__tests__/responsive-config-spec-parity.test.ts is deleted with them (its subject ResponsiveConfigSchema no longer exists in the spec; MobileResponsiveConfig stays finding(types): retiring MobileComponentConfig leaves MobileResponsiveConfig and GestureConfig as zero-consumer published types #7519's).
   5. The test(plugin-charts): serve ObjectChart's option-color probe from a double, not the network #4115 comment at mobile.ts:28 is answered, not deleted: the spec no longer exports the name, so a local declaration is the only definition and cannot be mistaken for the spec's; the comment is rewritten to say exactly that and to cite objectstack#11027.

   ⛔ Not C: putting a renderer-only vocabulary back into @objectstack/spec where the engine has no reader is the declared-but-unenforced shape ADR-0049 removes, and it would reverse a merged, ruled retirement. ⛔ Not B: deleting a registered component's authorable input and four runtime readers on no evidence. Whether responsive-grid.columns is worth keeping at all (in-repo authoring: zero outside content/docs/guide/layout.md) is a separate liveness question — a finding for triage, ⛔ not folded here.

   Execution constraints, for the domain:ui seat

   • One PR, both halves — half 1 alone leaves the next spec consumption red (card's own point). Part of/Fixes per the seat's judgement; [finding] 响应式词汇的两个零消费表面:useResponsiveConfig 生产零调用者(page.components[].responsive 因此实际未生效)、MobileComponentConfig 零消费者 #4773 stays the measurement record and closes with it.
   • Clause-②: yes — @object-ui/types gains exported symbols (mechanical floor); contract-review tier; needs:contract-review on card and PR when the diff exists. Changeset: @object-ui/types and @object-ui/layout per their policies; @object-ui/mobile / @object-ui/core for the deletions.
   • Sequencing: objectui's next @objectstack/spec pin bump must carry Blocked-by: this card — the retirement is merged upstream and unreleased (spec at 17.2.0; tombstone cites RETIRED_DEFS_BY_MAJOR[18]), so this PR must land before that bump.
   • Serial constraints for the claim: finding(types): retiring MobileComponentConfig leaves MobileResponsiveConfig and GestureConfig as zero-consumer published types #7519 (MobileResponsiveConfig), any open PR on packages/types/src/mobile.ts or packages/layout/src/ResponsiveGrid.tsx.
   • Confidence gap carried into the dispatch: whether hotcrm / cloud import BreakpointName from @object-ui/types was not measured; the dev runs the consumer census (with a positive control) before deleting anything.

   Ruling authority: maintainer 2026-09-04 (batch #26). Related: #4773 · objectstack#11027 (35e94c96b) · #7519 · objectstack#4115.

   ---

   Generated by Claude Code
3. os-warren

   added

   pm:queue

   and removed

   needs-user-decision

   on Sep 4, 2026
4. claude commented on Sep 4, 2026

   

   claudeBot

   on Sep 4, 2026 – with Claude

   Contributor

   More actions

   Dev claim (os-dev subagent, re-dispatch after the fable 429)

   Session: session_01KbJQ1y1J12nZxYzFWhP8Q3
   Branch: claude/issue-7580-responsive-rehome
   Worktree: objectui-issue-7580
   Base: objectui origin/main = 00d3f09 (moved from the PM's 04:38Z reading of 77b3cf1; re-fetched, git rev-parse HEAD verified)
   Model: opus, per the tier amendment 5535740735 — Clause-②: yes is unchanged, draft PR only, landing is the PM's.

   Assignee is the PM's and stays os-sam; this comment is the identity bit. Card body read via REST and verified complete (6378 bytes, structure closed through the Provenance section, no sanitizer truncation).

   ---

   Generated by Claude Code
5. claude

   added

   needs:contract-review

   on Sep 4, 2026
6. claude commented on Sep 4, 2026

   

   claudeBot

   on Sep 4, 2026 – with Claude

   Contributor

   More actions

   os-dev-report

   {
     "issue": 7580,
     "status": "done",
     "branch": "claude/issue-7580-responsive-rehome",
     "pr": "https://github.com/objectstack-ai/objectui/pull/7598",
     "premise_still_valid": true,
     "summary": "Card body read via REST and verified COMPLETE (6378 bytes, structure closed through Provenance, no sanitizer truncation) — reporting that either way as instructed. All four ruling work-items were still undone on the real base: origin/main had moved 77b3cf1 -> 00d3f09 -> (merged) 4bb5e10, so every site was re-read rather than inherited. Both halves land in one PR. Half 2: BreakpointName declared locally in @object-ui/types (mobile.ts), BreakpointColumnMap in @object-ui/layout (ResponsiveGrid.tsx), both verbatim from the retired declarations; BreakpointOrderMap retired (ruling item 3); the objectstack#4115 comment ANSWERED and rewritten citing objectstack#11027 (ruling item 5). Half 1: useResponsiveConfig, ResponsiveProtocol and responsive-config-spec-parity.test.ts deleted. THREE SITES OUTSIDE THE RULING'S LIST, each in the card's own defect class and each of which would have left the bump red or a comment false, all named in the PR body: (1) types/src/index.ts:1349 re-exported ResponsiveConfig and BreakpointName STRAIGHT FROM THE SPEC as SpecResponsiveConfig / SpecBreakpointName — a second spec-import site nobody had named; dead once its two readers went, so dropped; (2) spec-ui-schema-reexports.test.ts carried three responsive deny-list rows under a ratchet that fails on rows the spec has retired — removed ahead of the ratchet firing, exactly as objectui#5716 handled ThemeModeSchema; (3) complex.ts:1037 asserted the ResponsiveConfig shape 'is NOT gone... which useResponsiveConfig really does read', which this PR falsifies — corrected in place, as was one sentence in mobile.ts's MobileResponsiveConfig note. KEY MECHANISM, measured not assumed: spec 17.2.0 PRE-dates the retirement and still exports both names, so a local declaration collides and check:spec-symbols goes red — I probed this before designing anything and captured the exact failure text. Held by two SELF-EXPIRING ALLOW entries in scripts/check-spec-symbol-derivation.mjs; that guard's ratchet 3 fails an entry excusing nothing, so the pin bump is MECHANICALLY FORCED to delete them. This is the repo's own precedent, not an invention: objectui#5716/#5668 did exactly this for the three theme document types under the same option A. Serial constraint honoured: PR #7577 has NOT landed (re-checked at merge time) and its file surface stays disjoint. Assignee untouched (os-sam). Draft, label not stripped, no ready flip, no enqueue, pin bump deliberately not done.",
     "tests": "All figures measured on the pushed head 2b63279 (after the final origin/main merge), quoting each gate's own verdict line; exit codes captured by redirect-then-read, never through a pipe. BUILD: pnpm --filter {types,core,layout,mobile} build green, plus the '^...' dependency-closure build first; rebuilt dist verified to actually carry the change (packages/types/dist/mobile.d.ts:63 'export type BreakpointName = ...', packages/layout/dist/ResponsiveGrid.d.ts:43 'export interface BreakpointColumnMap', and BreakpointOrderMap absent from layout/dist/index.d.ts) — so the type-checks below read a rebuilt .d.ts, not a cached one. TYPE-CHECK: all four green, script names echoed in the output ('type-check$ tsc --noEmit && tsc -p tsconfig.test.json'), so not a zero-match silent pass. VITEST (repo root, per this repo's guard): 238 files / 4255 tests, all passed. The edited gate script's OWN suites run separately from the derived families: scripts/__tests__/check-spec-symbol-derivation.test.ts + check-doc-component-types.test.ts = 2 files / 93 tests green. GATES: check:spec-symbols '1343 files scanned against 4959 spec export names; 20 declared dialects, 14 untriaged collisions' — 18 dialects before, so +2 is exactly my two entries and the untriaged count is unchanged (nothing pushed into DEBT); check:control-bytes 'OK (6223 tracked text files)'; check:doc-types and check:dist-completeness green. ESLINT on the 9 changed lintable files: 0 errors, 8 warnings, all no-explicit-any in complex.ts and IDENTICAL at baseline HEAD (verified by linting the HEAD blob through stdin) — my complex.ts change is comment-only, proved by diffing added non-comment lines (zero). NOTE the first eslint run showed 3 errors purely because I passed --no-inline-config, which disables the files' own sanctioned eslint-disable comments; re-run as CI runs it, clean. REVERSE VERIFICATION of the guard collision: mutation applied via perl -0pi, confirmed ON DISK before measuring (injected-line grep = 1, removed-import grep = 0, git hash-object differed from the HEAD blob), guard then failed with the exact quoted text; restored via 'git checkout HEAD -- packages/types/src/mobile.ts' inside a trap using an absolute repo root, and the restore PROVED by an empty 'git diff HEAD --stat' rather than by an exit code. DECLARED NARROWINGS, so they are not read as coverage: (a) check:readme-exports is NOT MEASURED locally rather than red — all 317 findings are the single prerequisite class 'type entry ./dist/index.d.ts is not on disk -- run pnpm build first' from packages outside my build closure, and ZERO name types/layout/mobile/core; the specific risk this diff creates (a README documenting a removed export) was measured directly instead and is zero across all 43 tracked READMEs. (b) Repo-wide pnpm lint and the full 43-package downstream type-check were not run locally (CI runs the farm); in their place the removal set was censused exhaustively BY NAME — SpecResponsiveConfig, SpecBreakpointName, useResponsiveConfig, ResolvedResponsiveState, BreakpointOrderMap, BreakpointKey, BREAKPOINT_VALUES, ResolvedResponsiveConfig, resolveResponsiveConfig, getVisibilityClasses, getColumnClasses, getOrderClasses, shouldHideAtBreakpoint — with ZERO live code references remaining (every survivor is a comment written here explaining the removal) against a BaseSchema control lit at 852. Type-aware linting is not enabled in eslint.config.js (no projectService/project/parserOptions), so this diff cannot move the verdict on any untouched file.",
     "mcp_calls": "1 — a single targeted search_code for the cross-repo consumer census; every other GitHub read and write went through the repo-scoped REST channel, which I probed open at the start (HTTP 200).",
     "open_questions": [
       {
         "question": "The hotcrm / cloud consumer census the ruling carried forward is STILL UNMEASURED, and I can now name the boundary rather than just report a gap: GET /repos/objectstack-ai/cloud and .../hotcrm both return HTTP 403 for this session, against HTTP 200 for objectui and objectstack. 403 not 404 — the repos exist and this session is not authorized. Because GitHub code search only indexes what the caller can read, the org-wide search STRUCTURALLY cannot see them, so its silence about those two repos is not evidence of absence and I have not reported it as any. What I did measure, with positive controls: org-wide 'BreakpointName org:objectstack-ai' returns 14 hits (9 objectui, matching the in-repo census file-for-file, which is what makes the query trustworthy rather than empty; 5 objectstack, all the retirement's own sources), and the objectstack sibling checkout consumes zero of it with @object-ui/ imports lit at 124 files. This does NOT gate the re-home, which is a no-op for consumers (same name, members and export sites — only provenance moved). It bites on exactly one item: BreakpointOrderMap's removal from @object-ui/layout is a published-surface removal, in-repo readers zero, external readers unmeasurable from this seat.",
         "options": [
           "A - accept as-is: the removal is ruling item 3, adjudicated, and the residual risk is one unread type; the changeset records it as a deliberate removal",
           "B - land the re-home now but defer only the BreakpointOrderMap deletion to a follow-up once someone with cloud/hotcrm access measures it",
           "C - grant this lane read access to cloud/hotcrm so the census can actually be run before landing"
         ],
         "recommendation": "A, with C as the durable fix. B buys almost nothing: BreakpointOrderMap has no read point even in the package that declared it, its own pre-existing comment says so, and it was published only because the retired ResponsiveConfigSchema paired it with the column map — so deferring it would keep a type alive whose entire reason for existing is already deleted upstream, and would split one ruled change across two PRs for a risk nobody can retire without C. C is worth raising with the maintainer independently of this card: this is the second time in one lane that a published-surface decision has had to be made on a census that a seat is structurally unable to run, and the 403 makes that permanent rather than incidental, not something a more careful dev can work around."
       }
     ],
     "out_of_scope_findings": [
       "filed as #7599: skills/objectui/rules/protocol.md:289 attributes the six-key breakpoint vocabulary to the spec's BreakpointColumnMapSchema, which objectstack#11027 retires — true on today's pin, false at the bump. Deliberately NOT fixed here: skills/** is a governed published surface with its own line ratchet, nothing breaks mechanically, and the doc describes the `grid` renderer (a different component from responsive-grid, and one that does not import the type at all), so it is an attribution error only. Filed unassigned with `finding`, no pm:queue. Dedupe: scanned 76 open + 100 closed finding-labelled issues, no duplicate.",
       "NOT filed, reported instead because it is already owned: at the pin bump, page-nav-misc-spec-parity.test.ts:608 ['ResponsiveConfig','MobileResponsiveConfig'] WILL fire — it asserts the spec still owns ResponsiveConfig, which is why MobileResponsiveConfig carries its renamed spelling, and the retirement makes that false. That is the tripwire working as designed (it has fired twice before, each time handled in the bump PR), and the triage it triggers is #7519's, which the ruling explicitly keeps out of this card. Left untouched deliberately: removing the row early would destroy the signal #7519 needs. Recorded in the PR body's Sequencing section so the bump is not surprised by it."
     ],
     "pm_mechanism_assumptions": {
       "a_spec_derived_unions_goes_red": "FALSIFIED — and this is the one I was asked to test. The suite does NOT go red: @object-ui/types type-check is green with the re-home in place, and the suite is genuinely in that compilation set (tsc -p tsconfig.test.json --listFiles puts it among 560 files), so this is a measurement and not an unmeasured pass. Reason: the case imports BreakpointName from '../mobile' — never from the spec — and asserts a WIDTH ('xs'|...|'2xl' satisfies BreakpointName) that the re-home preserves exactly. So neither of the two dispositions offered was right as stated. What expired is not the assertion but the COMMENT above it, which grouped the symbol under the objectstack#4115 'ledger burn-down' population: declarations proved equivalent to the spec's and then REPLACED BY A BINDING. This one went the other way. I restated it rather than deleting: the case keeps real enforcement (narrowing the now locally-owned union fails to compile instead of silently dropping a breakpoint), and the new comment records where the symbol goes next — at the bump it joins the suite's documented INVERSE population (names objectui uses that the spec does not have, asserted absent). That absence cannot be asserted yet because the pin still publishes the name, which is precisely what the self-expiring ALLOW entry covers.",
       "b_breakpointordermap_retires_cleanly": "CONFIRMED in-repo, with the external half unmeasurable (see open_questions). Zero references outside ResponsiveGrid.tsx and the CHANGELOGs; after the edit, zero live references anywhere, and it is absent from packages/layout/dist/index.d.ts.",
       "c_mobile_readers_need_no_change_beyond_import_source": "CONFIRMED, and stronger than stated — they needed NO change at all, not even to the import source. All four already imported BreakpointName from '@object-ui/types', not from the spec; only that package's internal provenance changed, so packages/mobile is untouched apart from deleting the dead hook's export line."
     }
   }

   ---

   Generated by Claude Code
7. hotlong

   removed

   needs:contract-review

   on Sep 4, 2026
8. hotlong commented on Sep 4, 2026

   

   hotlong

   on Sep 4, 2026 – with Claude

   Contributor

   More actions

   Contract review on PR #7598: PASS at head 2b63279 — director seat, summon 13, session_01WXyGTWPbbreqXow7Z2pZCk; the record is comment 5536189743 on the PR. Taken up because the domain:ui seat parked the PR (5536055641) as complete work it cannot review at tier. The cloud / hotcrm consumer census the lane could not run is measured there: zero readers of every removed name in both repos on origin/main, with positive controls — the dev's open question resolves as A (accept as-is). needs:contract-review cleared here and on the PR in this stroke (read-modify-write, read back), PR marked ready, auto-merge squash armed; this card closes with the merge and pm:dispatched goes with it.

   ---

   Generated by Claude Code
9. hotlong

   closed this as completedin #7598on Sep 4, 2026
10. akarma-synetal

    added a commit that references this issue on Sep 9, 2026

    fix(types,layout,mobile,core): re-home the breakpoint vocabulary, del…

    ...

    e62c44e